North Carolina Bootcamps

This post originally appeared on the Software Carpentry website.

Last month, we held two Software Carpentry workshops back-to-back in Durham, North Carolina, one at the National Evolutionary Synthesis Center and a second organized by Cliburn Chan for the biostats group at Duke just three days later. These workshops were taught by Jenny Bryan, Elliott Hauser, and myself with Karen Cranston teaching as well at NESCent. The workshops were both somewhat atypical: NESCent was a smaller workshop (only 11 participants) who work together and share an interest in evolutionary biology, and at both NESCent and Duke we presented in R rather than Python. Overall, both workshops were very successful, and I learned a lot, both about R and teaching a Software Carpentry workshop.

In addition to substituting R for Python, we altered the schedule from what I've seen in previous workshops, teaching R for the entire first day and everything else (shell, git, and Make) on day two. While this meant that Jenny had to present the entire day on day one, her material and instruction style easily held everyone's interest for the whole day. Jenny went at a good pace, moving from basic language essentials and how to interact with RStudio to more complex topics such as data aggregation using the plyr package and producing literate documents using knitr. The R sections, particularly the use of plyr, were consistent crowd favorites. I think the way we rearranged the schedule worked out very well—we avoided confusion by not switching between environments, and using only RStudio on day one allowed us to avoid installation issues. Anyone who had had trouble installing the prerequisites either stayed after on day one or came early on day two for help. We ended up with no setup issues at NESCent and a fair number at Duke (possibly due to inadequate instructions on our boot camp page) but they were easily handled.

We were able, perhaps serendipitously, to present our material in a logically cohesive way that kept students engaged. Instead of every instructor bringing their own material, we used the same files for all sections. The code we wrote during day one was placed under version control on day two, and we wrote a Makefile together to automate our data analysis pipeline and produce an HTML document with code and figures. At NESCent, this wasn't planned ahead of time; I noticed that Jenny had designed her scripts in such a way that they were great candidates to incorporate into my Make tutorial, and so together we wrote one. There was clear interest in Make, so we built on this idea and replicated it at Duke with more preparation.

The number of people attending the NESCent workshop was low, which was in some ways both a pro and a con. On the plus side, we were able to pay more individual attention and Karen led a discussion on data management that was highly relevant to those attending. However, we seemed to run out of steam during the afternoon on day two, where we had planned to address whatever topics participants felt would be helpful but found a lack of enthusiastic suggestions. We corrected this at Duke by being prepared for a more structured presentation. We also found for both the shell and version control sections that our presentation needed to cater to specific tasks that a researcher just learning these skills would use and avoid more complicated topics such as forking and pull requests.

Teaching two workshops back-to-back gave us a chance to refine our approach and materials, and Jenny is looking into how best to incorporate the materials we designed into Software Carpentry for future R workshops to utilize.

NESCent feedback

Duke feedback